Other CPU Performance Patch (Up to 30% faster physics, up to 60% more FPS)

[hayarevn]

How to config for:
i5 13500Hx
Nvdia rtx 4060 8G
Ram 16G
Thanks.

 

[ToeJam]

How to unlimit the 16 cpu limitation?

I had chance to try VaM on many cores desktop (but very old...).
This patch seems to improve fps well. But it seems still limited only for 16cpu of 2*22cpu cores.(2*44 threads)

SkinMeshPartDLL.ini:
-----
[threads]
affinity=1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79,81,83,85,87

[profiler]
enabled=0
-----

boot.config:
-----
gfx-enable-gfx-jobs=1
gfx-enable-native-gfx-jobs=1
wait-for-native-debugger=0
-----

 

[ToeJam]

The affinity options seems to set only the first 32 cpus. (So, 16 cpus for skipping HTs at my previous post case)

If BIOS setting disables HT & NUMA, can set 32 cpu cores affinity. But...
- with no affinity (1-44cpu), got lower fps even use all cpu cores.
- 1-22 cpu, still lower...
- 1-16 cpu, still lower... (Even if NUMA enabled )

Yes, the previous case (enable HT, 16 cores skipping HTs) was the best case for now. Might not so needed. Thanks.

 

[turtlebackgoofy]

The affinity options seems to set only the first 32 cpus. (So, 16 cpus for skipping HTs at my previous post case)

If BIOS setting disables HT & NUMA, can set 32 cpu cores affinity. But...
- with no affinity (1-44cpu), got lower fps even use all cpu cores.
- 1-22 cpu, still lower...
- 1-16 cpu, still lower... (Even if NUMA enabled )

Yes, the previous case (enable HT, 16 cores skipping HTs) was the best case for now. Might not so needed. Thanks.

windows only supports a 32bit mask, so 32 cores. Beyond that you need to use numa and a different API. I didnt expect anyone to have more than 32 cores to use vam since those CPUs have horrible single thread performance.

 

[mostvanvege]

windows only supports a 32bit mask, so 32 cores. Beyond that you need to use numa and a different API. I didnt expect anyone to have more than 32 cores to use vam since those CPUs have horrible single thread performance.

Really? So thats why desktop cpus stopped at 32 thread?! (AMD x950 with 2 ccd maxed at 16c 32t and Intel RPL 13900 was also 8p+16e conf which were 32t..)

 

[turtlebackgoofy]

Really? So thats why desktop cpus stopped at 32 thread?! (AMD x950 with 2 ccd maxed at 16c 32t and Intel RPL 13900 was also 8p+16e conf which were 32t..)

no, since windows vista you can use more than 32 threads but you need to use another API to assign threads to cores above 32. Also some CPUs are on different NUMA cores and you need to be careful to not split workloads between different NUMAs. I didnt bother to use the other API since I cant test it with my CPU.

 

[mostvanvege]

no, since windows vista you can use more than 32 threads but you need to use another API to assign threads to cores above 32. Also some CPUs are on different NUMA cores and you need to be careful to not split workloads between different NUMAs. I didnt bother to use the other API since I cant test it with my CPU.

I see, thanks for explanation! Tough if it means additional programming I still feel cpu manufacturers somehow tried to avoid complications due to thread count.

 

[StyleMavin]

If you are trying to use Virtual Desktop through a shared wifi, results are going to be from bad to nauseating. You need a dedicated router connected to your pc with an ethernet cable, with the oculus connecting wirelessly to that. I had already bought a second router since our old internet provider only provided a gateway and when connecting that way performance is great and the extra controls come in handy too.
If I forget to switch the connection in the quest and connect through the house wifi lag is insane, no matter if I am on VaM or any other program

Update. Browser Assist fixed the problem by offloading VAR files. I don't need Virtual Desktop to enjoy decent frame rates now.

 

[ToeJam]

windows only supports a 32bit mask, so 32 cores. Beyond that you need to use numa and a different API. I didnt expect anyone to have more than 32 cores to use vam since those CPUs have horrible single thread performance.

BTW, Why 32 bit? is it 64bit mask?
SetProcessAffinityMask has parameter "[in] DWORD_PTR dwProcessAffinityMask" and this is not as pointer, but data as it is.
And DWORD_PTR should be 64 bit data.

 

[mediamac1]

for an i9-14900k with 4090 I chose the same settings for VR via Quest3/Wifi6e. Why is a weaker dimensioning recommended for this?

I increased it and find the quality better.

Can anyone explain the situation/optimal settings here?

[threads]
computeColliders=8
skinmeshPart=8
applyMorphs=8
skinmeshPartMaxPerChar=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15

[threadsVR]
computeColliders=8
skinmeshPart=8
applyMorphs=8
skinmeshPartMaxPerChar=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15

 

[turtlebackgoofy]

BTW, Why 32 bit? is it 64bit mask?
SetProcessAffinityMask has parameter "[in] DWORD_PTR dwProcessAffinityMask" and this is not as pointer, but data as it is.
And DWORD_PTR should be 64 bit data.

you are correct, documentation even states that you can assign 64 CPU cores. Beyond that you need to put the cores into groups and then assign groups of up to 64 cores each to threads.

 

[Borogravian]

A security patch for Virt-a-Mate came out today, 1.22.0.6 . I'm assuming installing this performance patch will overwrite the security changes?

If so, hoping to see a version of this supporting the 1.22.0.6 version.

 

[Sally Whitemane]

... hoping to see a version of this supporting the 1.22.0.6 version.

Assuming 'they' means turtlebackgoofy I'm optimistic. Active user on the hub too.

Now everybody on your knees!
Pray to the CPU performance-patch-Overlord!

(Intentionally keeping my mouth shut about old version dll-swap-shenanigans. Could undermine the security and lead to many other problems. Better wait for proper compatible CPU perf.-patch.)

 

[meshedvr]

(Intentionally keeping my mouth shut about old version dll-swap-shenanigans. Could undermine the security and lead to many other problems. Better wait for proper compatible CPU perf.-patch.)

Yeah I think the proper thing to do after this resource is updated is to remove all the older versions since they have the security flaw.

 

[Chokaphi]

Yeah I think the proper thing to do after this resource is updated is to remove all the older versions since they have the security flaw.

Did the author ever give the code changes he made to you? Or is it only provided the patched dll?

 

[turtlebackgoofy]

since I dont have the source code of vam itself I just hotpatched the dlls and thats actually how I keep my code changes haha

 

[Makimax]

is this patch ok with the 1.22.0.6 patch ?

 

[turtlebackgoofy]

is this patch ok with the 1.22.0.6 patch ?

no, this patch will just revert back to the old vam version, dont use it until I update it

 

[Case]

Pray to the CPU performance-patch-Overlord!

"Pie Turtle Domine, dona eis requiem ..."

Spoiler

 

[Damarmau]

Waiting on this before I update VaM. My ancient 4770k just can't go back to running VaM without it.

 

[lblg]

1.22.0.7 update! wow

 

[sdfg3456]

1.22.0.7 update! wow

what's wow about it?

 

[Hun73rdk]

Can we get a update for latest version of vam ??

 

[Zihobi]

what's wow about it?

He was trying to be subtle

 

[atani]

Can we get a update for latest version of vam ??

There is no need to ask for updates, that will not help in any way and instead only annoys or makes the creator feel bad.
The creator is aware of the VaM updates and has a life of his own, any update will be at the creator's convenience.